The impression I've got from this mailing list is they are
intentional design decisions to enforce "correctness" of the cluster.
My understanding of cluster (ever since the VAX) is that it's a fault-tolerance
mechanism and that was originally one of the major selling points of these hypervisor
orchestration suites like vSphere, RHV etc. Since that doesn't quite work with local
storage, I understand why they left it out for live migration. AFAIK it works just fine
for VMs that are down or disks that are not attached.